The part of dataset in here https://drive.google.com/drive/folders/1cZEUnyF3jn0UnQNrhlZCkVjb54XzRTBd?hl=zh-CN Note that this code is heavily borrowed from RandLA-Net (https://github.com/QingyongHu/RandLA-Net).

(1) Setup

This code has been tested with Python 3.5, Tensorflow 1.15, CUDA 10.0 and cuDNN 7.4.1 on Ubuntu 16.04.

• Setup python environment

  conda create -n randlanet python=3.5
  source activate randlanet
  pip install -r helper_requirements.txt
  sh compile_op.sh

(2) BF

/data/BF.

• Preparing the dataset:

  python utils/data_prepare_BF.py

• Start 6-fold cross validation:

  sh jobs_6_fold_cv_bf.sh

• Move all the generated results (*.ply) in /test folder to /data/BF/results, calculate the final mean IoU results:

  python utils/6_fold_cv.py

  Citation

  If you find our work useful in your research, please consider citing:

  @article{su2022dla, title={DLA-Net: Learning dual local attention features for semantic segmentation of large-scale building facade point clouds}, author={Yanfei Su and Weiquan Liu and Zhimin Yuan and Ming Cheng and Zhihong Zhang and Xuelun Shen and Cheng Wang}, journal={Pattern Recognition}, volume = {123}, pages = {108372}, year = {2022}, issn = {0031-3203}, doi = {https://doi.org/10.1016/j.patcog.2021.108372}, url = {https://www.sciencedirect.com/science/article/pii/S0031320321005525}, }